Enhancing Factor Timing Predictions: The Role of Economic Structure and Deep Learning
quantseeker @ twitter
orig
A new paper emphasizes the importance of incorporating economic structure and time series dynamics in factor timing models, significantly enhancing predictive accuracy. Critical variables identified for effective factor timing include tail risk, price trends, and leverage. Additionally, I've launched a newsletter aimed at providing weekly insights into quant research focused on investing, macroeconomic trends, and trading strategies.
@quant_feed
quantseeker @ twitter
orig
A new paper emphasizes the importance of incorporating economic structure and time series dynamics in factor timing models, significantly enhancing predictive accuracy. Critical variables identified for effective factor timing include tail risk, price trends, and leverage. Additionally, I've launched a newsletter aimed at providing weekly insights into quant research focused on investing, macroeconomic trends, and trading strategies.
@quant_feed
Navigating Low Liquidity: Understanding Market Vulnerabilities During Seasonal Calm
Ksidiii @ twitter
orig
We're entering a quiet seasonal period in the market, which heightens susceptibility to irrational headlines and narratives amid light liquidity. This phenomenon is a recurring theme each summer. While volatility may seem clustered at low levels, don't underestimate the potential for sudden spikes in volatility, as seen in past summers.
@quant_feed
Ksidiii @ twitter
orig
We're entering a quiet seasonal period in the market, which heightens susceptibility to irrational headlines and narratives amid light liquidity. This phenomenon is a recurring theme each summer. While volatility may seem clustered at low levels, don't underestimate the potential for sudden spikes in volatility, as seen in past summers.
@quant_feed
Rethinking Predictive Models: Why Past Performance Isn't an Indicator of Future Success
Citrini7 @ twitter
orig
Leaders who drive progress do not limit their vision based on past performance; this narrow perspective stifles innovation. It's amusing how articles elevate firms like Goldman Sachs for their focus on profitability while ignoring their true aim of attracting trading volume. Claims about the necessity of infrastructure often get debunked by innovation, showcasing the dynamic nature of industry. The presence of a Fed put illustrates how markets can remain buoyant despite skepticism. Amy embodies the spirit of resilience and adaptability we need to embrace.
@quant_feed
Citrini7 @ twitter
orig
Leaders who drive progress do not limit their vision based on past performance; this narrow perspective stifles innovation. It's amusing how articles elevate firms like Goldman Sachs for their focus on profitability while ignoring their true aim of attracting trading volume. Claims about the necessity of infrastructure often get debunked by innovation, showcasing the dynamic nature of industry. The presence of a Fed put illustrates how markets can remain buoyant despite skepticism. Amy embodies the spirit of resilience and adaptability we need to embrace.
@quant_feed
Exploring the Nature of Equity Factors: Mispricing vs. Risk in Quantitative Finance
quantseeker @ twitter
orig
Hundreds of equity factors exist, but the debate remains whether they indicate mispricing or just risk. Frey’s analysis suggests that at least 40% of factors signal mispricing, emphasizing that many factors are essentially the mechanisms through which prices adjust back to fundamental values. This highlights the crucial interplay between price dynamics and fundamental analysis in the investment landscape. Staying updated on emerging research in investing, macroeconomic trends, and trading can illuminate these concepts further, enhancing our understanding of market behavior.
@quant_feed
quantseeker @ twitter
orig
Hundreds of equity factors exist, but the debate remains whether they indicate mispricing or just risk. Frey’s analysis suggests that at least 40% of factors signal mispricing, emphasizing that many factors are essentially the mechanisms through which prices adjust back to fundamental values. This highlights the crucial interplay between price dynamics and fundamental analysis in the investment landscape. Staying updated on emerging research in investing, macroeconomic trends, and trading can illuminate these concepts further, enhancing our understanding of market behavior.
@quant_feed
Developing Predictive Models for Unusual Travel Patterns in Influencers
PtrPomorski @ twitter
orig
When choosing a model for predictive analysis in finance, it's critical to rethink the conventional preference for LSTMs often suggested by platforms like Kaggle. I advocate starting with regularized linear models and Generalized Additive Models (GAMs), such as ridge, lasso, and elastic net. These models are straightforward and have effective applications in high-frequency trading, despite the underlying assumption of independence being questionable in financial contexts.
Transitioning to non-linear methods, tree-based algorithms like LGBM often outperform others in this arena. They fit the interconnected nature of financial data well but come with complexities that can lead to overfitting and calibration challenges, particularly in calculating confidence intervals. Despite these drawbacks, tree models are generally more suited for lower frequency data and require less preprocessing compared to their linear counterparts.
Lastly, while neural networks and their combinations with GAMs (NAMs) are gaining traction, their utility is mostly seen in deep hedging scenarios. They offer significant out-of-sample accuracy given sufficient data and compute resources, but their complexity can hinder analysis and feature importance evaluation. Remember, starting with simple linear models not only provides a good foundation but also a clear benchmark for assessing the additional insights drawn from more intricate algorithms.
@quant_feed
PtrPomorski @ twitter
orig
When choosing a model for predictive analysis in finance, it's critical to rethink the conventional preference for LSTMs often suggested by platforms like Kaggle. I advocate starting with regularized linear models and Generalized Additive Models (GAMs), such as ridge, lasso, and elastic net. These models are straightforward and have effective applications in high-frequency trading, despite the underlying assumption of independence being questionable in financial contexts.
Transitioning to non-linear methods, tree-based algorithms like LGBM often outperform others in this arena. They fit the interconnected nature of financial data well but come with complexities that can lead to overfitting and calibration challenges, particularly in calculating confidence intervals. Despite these drawbacks, tree models are generally more suited for lower frequency data and require less preprocessing compared to their linear counterparts.
Lastly, while neural networks and their combinations with GAMs (NAMs) are gaining traction, their utility is mostly seen in deep hedging scenarios. They offer significant out-of-sample accuracy given sufficient data and compute resources, but their complexity can hinder analysis and feature importance evaluation. Remember, starting with simple linear models not only provides a good foundation but also a clear benchmark for assessing the additional insights drawn from more intricate algorithms.
@quant_feed
Optimizing Fill Performance While Addressing Drawdown Challenges
BeatzXBT @ twitter
orig
Progress in reducing drawdown spikes is evident, but there's still room for improvement in fill quality. The PnL graph remains relatively stable; I've focused on enhancing its smoothness following recent changes. Exploring Grafana and TimescaleDB for live monitoring is on my radar, though Streamlit is also an option, I’m leaning towards mastering TimescaleDB for efficiency. I've made some tweaks to the quoting strategy, but the core approach remains simple. While I could dig deeper into data trends, my current bandwidth is limited.
@quant_feed
BeatzXBT @ twitter
orig
Progress in reducing drawdown spikes is evident, but there's still room for improvement in fill quality. The PnL graph remains relatively stable; I've focused on enhancing its smoothness following recent changes. Exploring Grafana and TimescaleDB for live monitoring is on my radar, though Streamlit is also an option, I’m leaning towards mastering TimescaleDB for efficiency. I've made some tweaks to the quoting strategy, but the core approach remains simple. While I could dig deeper into data trends, my current bandwidth is limited.
@quant_feed
Navigating Career Pathways in Top Quant Funds: A Reality Check for Aspiring Candidates
PtrPomorski @ twitter
orig
It appears that top funds like Jane Street and Citadel are currently the gold standard for financial careers. However, accessing these opportunities is extremely challenging if you’re not from a target school—effectively, you’re out of the running. For those from target schools, the interview process is a brutal filter where the overwhelming majority of candidates, about 99%, may not succeed.
Interestingly, starting your career elsewhere as a junior and then aiming to transition back into these elite firms may be a more viable strategy. There's been a lot of chatter about the effectiveness of platforms like LinkedIn for job hunting, with mixed reviews on whether anyone has actually secured positions through it. This recruitment trend seems to perpetuate an elitist atmosphere, creating barriers that favor a select few.
@quant_feed
PtrPomorski @ twitter
orig
It appears that top funds like Jane Street and Citadel are currently the gold standard for financial careers. However, accessing these opportunities is extremely challenging if you’re not from a target school—effectively, you’re out of the running. For those from target schools, the interview process is a brutal filter where the overwhelming majority of candidates, about 99%, may not succeed.
Interestingly, starting your career elsewhere as a junior and then aiming to transition back into these elite firms may be a more viable strategy. There's been a lot of chatter about the effectiveness of platforms like LinkedIn for job hunting, with mixed reviews on whether anyone has actually secured positions through it. This recruitment trend seems to perpetuate an elitist atmosphere, creating barriers that favor a select few.
@quant_feed
Exploring Performance: Hyper vs. Reqwest in Rust Development
Dub0x3A @ twitter
orig
Hyper outperforms Reqwest in Rust, showing an ~18% speed advantage in a test with a simple local HTTP/1 server. If you're aiming for micro-optimizations, Hyper's lower-level client is worth considering. Kudos to @seanmonstar and contributors for their solid work on these libraries, and shoutout to @fasterthanlime for the helpful article. While Hyper is efficient and adopted by major players like AWS, Discord, and Cloudflare, I recommend most users stick with Reqwest for its ease of use and comprehensive handling of complexities like TCP screams, ALPN negotiation, and TLS.
@quant_feed
Dub0x3A @ twitter
orig
Hyper outperforms Reqwest in Rust, showing an ~18% speed advantage in a test with a simple local HTTP/1 server. If you're aiming for micro-optimizations, Hyper's lower-level client is worth considering. Kudos to @seanmonstar and contributors for their solid work on these libraries, and shoutout to @fasterthanlime for the helpful article. While Hyper is efficient and adopted by major players like AWS, Discord, and Cloudflare, I recommend most users stick with Reqwest for its ease of use and comprehensive handling of complexities like TCP screams, ALPN negotiation, and TLS.
@quant_feed
Choosing Quality Over Quantity: The Value of Experienced Insights in Quantitative Finance Literature
macrocephalopod @ twitter
orig
Investing in quality educational resources is crucial. Instead of wasting $300 a year on a subpar quant trading substack, consider allocating that budget to 4-6 professional books rich with practical insights gained from decades of industry experience. While some subscriptions might promise to elevate your trading skills, the harsh reality is that the majority won’t translate to actual profits. There’s a fine line between creating content and recommending it; if it lacks depth, I wouldn’t promote it authentically. Always prioritize resources that deliver substantial, actionable knowledge grounded in real-world experience.
@quant_feed
macrocephalopod @ twitter
orig
Investing in quality educational resources is crucial. Instead of wasting $300 a year on a subpar quant trading substack, consider allocating that budget to 4-6 professional books rich with practical insights gained from decades of industry experience. While some subscriptions might promise to elevate your trading skills, the harsh reality is that the majority won’t translate to actual profits. There’s a fine line between creating content and recommending it; if it lacks depth, I wouldn’t promote it authentically. Always prioritize resources that deliver substantial, actionable knowledge grounded in real-world experience.
@quant_feed
Exploring the Concept of Portable Alpha in Today's Investment Strategies
macrocephalopod @ twitter
orig
When engaging in fast trading, it's crucial to retain some fluff to maintain your edge, rather than completely hedging it out. You can often treat trades independently, evaluating whether the delta is increasing or decreasing. For very short-term trades (under five minutes), your size will typically be constrained by the available liquidity on the bid/offer, so maximize your position within that limit.
@quant_feed
macrocephalopod @ twitter
orig
When engaging in fast trading, it's crucial to retain some fluff to maintain your edge, rather than completely hedging it out. You can often treat trades independently, evaluating whether the delta is increasing or decreasing. For very short-term trades (under five minutes), your size will typically be constrained by the available liquidity on the bid/offer, so maximize your position within that limit.
@quant_feed
Exploring the Disparities Between Trading Fees and Tick Sizes in Market Microstructure
ltrd_ @ twitter
orig
I delved into market microstructure by analyzing BTCUSDT and CHZUSDT spots on Binance and BTCUSD perpetuals on ByBit. The study revealed significant differences in taker fees and tick sizes across these instruments. Notably, ByBit imposes a much higher taker fee, while CHZUSDT has a larger tick size compared to BTCUSDT on Binance.
Instantaneous market impact was assessed by observing how many levels in the order book vanished following market orders. For Binance's BTCUSDT, nearly 500 levels disappeared after a trade worth $3.6M, demonstrating substantial market impact. Conversely, CHZUSDT showed a maximum of only 8 levels disappearing, with negligible occurrences of high impact trades.
These findings highlight the critical role of optimal order placement for market makers. There's a vital trade-off between profit per trade and fill probability, complicating the strategy, especially when large market orders can consume numerous orders and affect trade execution. Furthermore, trading in markets with thin order books can introduce significant noise, rendering volume imbalance indicators unreliable and challenging to compare across exchanges.
It’s essential for traders, particularly market makers, to denoise such data to enhance strategy accuracy and to adapt their order placement logic depending on tick sizes. This nuanced understanding can lead to more informed trading decisions in varied market environments.
@quant_feed
ltrd_ @ twitter
orig
I delved into market microstructure by analyzing BTCUSDT and CHZUSDT spots on Binance and BTCUSD perpetuals on ByBit. The study revealed significant differences in taker fees and tick sizes across these instruments. Notably, ByBit imposes a much higher taker fee, while CHZUSDT has a larger tick size compared to BTCUSDT on Binance.
Instantaneous market impact was assessed by observing how many levels in the order book vanished following market orders. For Binance's BTCUSDT, nearly 500 levels disappeared after a trade worth $3.6M, demonstrating substantial market impact. Conversely, CHZUSDT showed a maximum of only 8 levels disappearing, with negligible occurrences of high impact trades.
These findings highlight the critical role of optimal order placement for market makers. There's a vital trade-off between profit per trade and fill probability, complicating the strategy, especially when large market orders can consume numerous orders and affect trade execution. Furthermore, trading in markets with thin order books can introduce significant noise, rendering volume imbalance indicators unreliable and challenging to compare across exchanges.
It’s essential for traders, particularly market makers, to denoise such data to enhance strategy accuracy and to adapt their order placement logic depending on tick sizes. This nuanced understanding can lead to more informed trading decisions in varied market environments.
@quant_feed
👍3
Exploring Optimal Trading Strategies and the Costs of Suboptimal Decisions in Market Impact Analysis
0xfdf @ twitter
orig
I recently reviewed two companion papers advancing our understanding of market impact: "The Cost of Misspecifying Market Impact" and "Trading with Concave Price Impact and Impact Decay." The key contributions are explicit optimal trading rules that incorporate nonlinear impact and decay, along with quantifying the "costs" of misestimating market impact.
There's an intrinsic conflict between expected returns (alpha signals) and implementation costs arising from market impact. The literature usually focuses on portfolio degradation from alpha estimation errors but overlooks the consequences of misspecifying market impact. Market impact is modeled using two critical parameters: concavity and decay, where concavity indicates that larger orders are proportionally cheaper, and decay captures the time taken for price reversion.
The classical square root model supports linear impact but lacks decay. The new model integrates the impact dynamics with a martingale approach, allowing for closed-form solutions, even with time-varying parameters. Optimal trading strategies derived from this model indicate that underestimating impact risks aggressive trading, which can erode profits, while overestimating leads to overly timid trades without major detriment.
A pivotal aspect is the model's immunity to price manipulation, establishing a no-arbitrage condition that confirms its robustness against inconsistencies. Empirical fitting of the model with a substantial dataset supports a fitted concavity around 0.48 and an optimal decay horizon near 0.2 days, challenging traditional models like the square root law, especially outside typical trading volumes.
While I prefer the Almgren model for its simplicity and empirical clarity in portfolio optimization, the insights and methodology emerging from this current research in stochastic control are promising and could drive future advancements in algorithmic trading strategies.
@quant_feed
0xfdf @ twitter
orig
I recently reviewed two companion papers advancing our understanding of market impact: "The Cost of Misspecifying Market Impact" and "Trading with Concave Price Impact and Impact Decay." The key contributions are explicit optimal trading rules that incorporate nonlinear impact and decay, along with quantifying the "costs" of misestimating market impact.
There's an intrinsic conflict between expected returns (alpha signals) and implementation costs arising from market impact. The literature usually focuses on portfolio degradation from alpha estimation errors but overlooks the consequences of misspecifying market impact. Market impact is modeled using two critical parameters: concavity and decay, where concavity indicates that larger orders are proportionally cheaper, and decay captures the time taken for price reversion.
The classical square root model supports linear impact but lacks decay. The new model integrates the impact dynamics with a martingale approach, allowing for closed-form solutions, even with time-varying parameters. Optimal trading strategies derived from this model indicate that underestimating impact risks aggressive trading, which can erode profits, while overestimating leads to overly timid trades without major detriment.
A pivotal aspect is the model's immunity to price manipulation, establishing a no-arbitrage condition that confirms its robustness against inconsistencies. Empirical fitting of the model with a substantial dataset supports a fitted concavity around 0.48 and an optimal decay horizon near 0.2 days, challenging traditional models like the square root law, especially outside typical trading volumes.
While I prefer the Almgren model for its simplicity and empirical clarity in portfolio optimization, the insights and methodology emerging from this current research in stochastic control are promising and could drive future advancements in algorithmic trading strategies.
@quant_feed
Understanding the Recent Surge in Japanese Yields and Yen Strength
MacroAlf @ twitter
orig
The Bank of Japan has raised interest rates again, establishing a short-term policy range around 0.25%, while also significantly reducing its government bond purchases over the next two years. Governor Ueda is optimistic about achieving a stable 2% inflation target, bolstered by significant wage increases secured by major labor unions. Although real rates remain negative, the BoJ's hawkish stance indicates further hikes could follow if core inflation approaches the target.
The market response has been sharp, with rising Japanese yields and a strengthening yen. Japanese investors, pivotal as global capital exporters, control over $1 trillion in US Treasuries and $0.5 trillion in EUR bonds. However, the current expense of FX hedging and the inverted US yield curve are making US Treasuries less attractive relative to Japanese government bonds, leading to potential capital allocation shifts.
As these changes manifest, the implications for global markets are significant, particularly in bond and stock allocations influenced by Japanese capital flows. Japan's monetary policy adjustments are not just local phenomena; they resonate throughout the global financial landscape.
@quant_feed
MacroAlf @ twitter
orig
The Bank of Japan has raised interest rates again, establishing a short-term policy range around 0.25%, while also significantly reducing its government bond purchases over the next two years. Governor Ueda is optimistic about achieving a stable 2% inflation target, bolstered by significant wage increases secured by major labor unions. Although real rates remain negative, the BoJ's hawkish stance indicates further hikes could follow if core inflation approaches the target.
The market response has been sharp, with rising Japanese yields and a strengthening yen. Japanese investors, pivotal as global capital exporters, control over $1 trillion in US Treasuries and $0.5 trillion in EUR bonds. However, the current expense of FX hedging and the inverted US yield curve are making US Treasuries less attractive relative to Japanese government bonds, leading to potential capital allocation shifts.
As these changes manifest, the implications for global markets are significant, particularly in bond and stock allocations influenced by Japanese capital flows. Japan's monetary policy adjustments are not just local phenomena; they resonate throughout the global financial landscape.
@quant_feed
Understanding Correlation in Quant Trading: Defining a "Good" Signal-Return Relationship
macrocephalopod @ twitter
orig
Correlation between your signal and future returns is crucial in quant trading. A key metric to consider is establishing what constitutes a "good" correlation. Using a simple model where future returns are normally distributed, we can derive beta based on correlation, volatility, and forecast horizon. This understanding implies that if trading costs are considered, identifying unprofitable signals becomes much simpler.
For instance, considering a stock with 3% daily volatility and 5 bps trading costs, we can expect to easily find signals with a correlation of 0.5%. If we engage factor hedging, we could halve the volatility but double our costs, aiming for a minimum correlation of 2% with idiosyncratic returns. In a different scenario, like predicting a 1-minute FX return with 0.2 bps trading costs and 0.3% daily volatility, we anticipate alphas with a much higher correlation of 8.5%.
These values represent baseline correlations; profitable trading requires exceeding these thresholds. A rule of thumb suggests that a correlation of 1.5 times the minimum allows for trades in about 5% of periods, while 2 times the minimum is considered very good. This framework is a practical way to extend the law of active management, emphasizing that realistic parameters will generally yield correlations less than 1.
@quant_feed
macrocephalopod @ twitter
orig
Correlation between your signal and future returns is crucial in quant trading. A key metric to consider is establishing what constitutes a "good" correlation. Using a simple model where future returns are normally distributed, we can derive beta based on correlation, volatility, and forecast horizon. This understanding implies that if trading costs are considered, identifying unprofitable signals becomes much simpler.
For instance, considering a stock with 3% daily volatility and 5 bps trading costs, we can expect to easily find signals with a correlation of 0.5%. If we engage factor hedging, we could halve the volatility but double our costs, aiming for a minimum correlation of 2% with idiosyncratic returns. In a different scenario, like predicting a 1-minute FX return with 0.2 bps trading costs and 0.3% daily volatility, we anticipate alphas with a much higher correlation of 8.5%.
These values represent baseline correlations; profitable trading requires exceeding these thresholds. A rule of thumb suggests that a correlation of 1.5 times the minimum allows for trades in about 5% of periods, while 2 times the minimum is considered very good. This framework is a practical way to extend the law of active management, emphasizing that realistic parameters will generally yield correlations less than 1.
@quant_feed
Announcing a New Book on Quantitative Investing: "The Element of Quantitative Investing" Set for April 2025 Release
__paleologo @ twitter
orig
I'm wrapping up my latest book draft, titled "The Element of Quantitative Investing," set for Wiley in April 2025. This book is the culmination of my thoughts on quantitative investing that I’ve wanted to express for a while. It covers essential topics like modeling returns and risk, backtesting alpha, portfolio construction, intertemporal and Kelly criteria, strategy execution, and performance analysis.
I’ve streamlined the content, cutting out unnecessary material to keep it under 500 pages, which makes it a focused read rather than a reference or thesis. The draft is close to completion but still contains typographical errors and needs refinement. The goal is for readers to grasp the core concepts to a level suitable for reimplementation, so the text will contain practical explanations and examples. I deliberately avoid any math newer than 50 years, focusing on timeless concepts relevant for the future.
Feedback is welcomed—email me corrections or comments with "EQI" in the subject line. The draft and related materials are accessible through the provided links, with all chapters currently in a readable state. Notably, I'm keeping the chapters on execution and signal fusion under wraps for now.
@quant_feed
__paleologo @ twitter
orig
I'm wrapping up my latest book draft, titled "The Element of Quantitative Investing," set for Wiley in April 2025. This book is the culmination of my thoughts on quantitative investing that I’ve wanted to express for a while. It covers essential topics like modeling returns and risk, backtesting alpha, portfolio construction, intertemporal and Kelly criteria, strategy execution, and performance analysis.
I’ve streamlined the content, cutting out unnecessary material to keep it under 500 pages, which makes it a focused read rather than a reference or thesis. The draft is close to completion but still contains typographical errors and needs refinement. The goal is for readers to grasp the core concepts to a level suitable for reimplementation, so the text will contain practical explanations and examples. I deliberately avoid any math newer than 50 years, focusing on timeless concepts relevant for the future.
Feedback is welcomed—email me corrections or comments with "EQI" in the subject line. The draft and related materials are accessible through the provided links, with all chapters currently in a readable state. Notably, I'm keeping the chapters on execution and signal fusion under wraps for now.
@quant_feed
Essential Questions for Assessing Linear Algebra and Regression Insights in Quant Interviews
macrocephalopod @ twitter
orig
In discussions about quant interviews, there's a significant focus on assessing fundamental concepts rather than practical applications like alpha research or portfolio construction. Specifically, probing candidates' grasp of linear algebra, regression, covariance matrix estimation, and dimension reduction can yield insights into their foundational knowledge. Engaging with others on Twitter reveals a diverse range of preferred questions that target these essentials. Some responses suggest skepticism about the relevance of certain questions, while others emphasize the importance of evaluating a candidate’s core understanding. It's clear that interviewing strategies should emphasize conceptual clarity over rote knowledge.
@quant_feed
macrocephalopod @ twitter
orig
In discussions about quant interviews, there's a significant focus on assessing fundamental concepts rather than practical applications like alpha research or portfolio construction. Specifically, probing candidates' grasp of linear algebra, regression, covariance matrix estimation, and dimension reduction can yield insights into their foundational knowledge. Engaging with others on Twitter reveals a diverse range of preferred questions that target these essentials. Some responses suggest skepticism about the relevance of certain questions, while others emphasize the importance of evaluating a candidate’s core understanding. It's clear that interviewing strategies should emphasize conceptual clarity over rote knowledge.
@quant_feed
Navigating PostgreSQL Connectivity in Python: A Guide to psycopg3 and psycopg2
ryxcommar @ twitter
orig
Connecting to Postgres in Python is straightforward with psycopg3, the latest driver. Despite its advancements, psycopg2 remains widely used, especially on Mac where you need to install psycopg2-binary. However, for a cleaner setup, your requirements.txt should reference the non-binary version. The consensus in the community seems to lean towards momentum for upgrades, but there’s still hesitance around updating database drivers, suggesting a general reluctance to change in established practices.
@quant_feed
ryxcommar @ twitter
orig
Connecting to Postgres in Python is straightforward with psycopg3, the latest driver. Despite its advancements, psycopg2 remains widely used, especially on Mac where you need to install psycopg2-binary. However, for a cleaner setup, your requirements.txt should reference the non-binary version. The consensus in the community seems to lean towards momentum for upgrades, but there’s still hesitance around updating database drivers, suggesting a general reluctance to change in established practices.
@quant_feed
Exploring the Evidence: Were US Stocks Truly "Lost" During Decades of Underperformance?
benjaminwfelix @ twitter
orig
Lost decades in US stocks are painful, especially when stocks trail one-month bills. Analyzing rolling 10-year periods from 1927 to 2023 reveals 145 instances classified as "lost decades," accounting for 14% of all periods. However, 74% of those instances saw the Dimensional US Small Cap Value Index outperforming US bills, with an average market return of +1.89%, resulting in a -2.33% premium over bills.
In contrast, US small cap value saw an average return of +6.45%, enhancing performance by 4.56% annualized above the market. This factor significantly mitigates the effects of fees and costs, which average around 2%.
Looking at Japan from 1987 to 2022, while the Fama/French Japan Market Index yielded only +2.81% annually, the Dimensional Japan Small Cap Value Index return surged to +8.11%. This indicates a consistent pattern where small cap value stocks have demonstrated resilience, despite their inherent risks and occasional underperformance.
Additionally, US large growth stocks have experienced more lost decades than the overall market, with 176 instances. During those periods, large growth underperformed the market by 0.66% annualized. The evidence suggests that small cap value can be a critical component in diversifying risks and enhancing returns, even during challenging market periods.
@quant_feed
benjaminwfelix @ twitter
orig
Lost decades in US stocks are painful, especially when stocks trail one-month bills. Analyzing rolling 10-year periods from 1927 to 2023 reveals 145 instances classified as "lost decades," accounting for 14% of all periods. However, 74% of those instances saw the Dimensional US Small Cap Value Index outperforming US bills, with an average market return of +1.89%, resulting in a -2.33% premium over bills.
In contrast, US small cap value saw an average return of +6.45%, enhancing performance by 4.56% annualized above the market. This factor significantly mitigates the effects of fees and costs, which average around 2%.
Looking at Japan from 1987 to 2022, while the Fama/French Japan Market Index yielded only +2.81% annually, the Dimensional Japan Small Cap Value Index return surged to +8.11%. This indicates a consistent pattern where small cap value stocks have demonstrated resilience, despite their inherent risks and occasional underperformance.
Additionally, US large growth stocks have experienced more lost decades than the overall market, with 176 instances. During those periods, large growth underperformed the market by 0.66% annualized. The evidence suggests that small cap value can be a critical component in diversifying risks and enhancing returns, even during challenging market periods.
@quant_feed
Exploring the Significance of Average Cross-Sectional R² in Factor Model Performance Metrics
__paleologo @ twitter
orig
The average cross-sectional R^2 is commonly cited as a primary performance metric for factor models, as noted by Barra and Axioma, but it falls short in several critical areas. First, R^2 does not effectively measure risk model performance, especially in the contexts of hedging or volatility prediction, nor does it correlate with alpha generation.
Furthermore, R^2 can be misleading due to its reliance on model complexity—it's inherently biased as it increases with the number of predictors. This can explain the proliferation of factors in commercial models. Additionally, data mining is a significant risk when assessing R^2, particularly given the limited historical data available for testing.
Despite these drawbacks, R^2 still provides an intuitive grasp of a factor's explanatory power. The challenge lies in rigorously linking population R^2 to genuine factor model performance across risk management and alpha generation, transcending the hurdles posed by finite sample considerations and multiple testing anomalies.
@quant_feed
__paleologo @ twitter
orig
The average cross-sectional R^2 is commonly cited as a primary performance metric for factor models, as noted by Barra and Axioma, but it falls short in several critical areas. First, R^2 does not effectively measure risk model performance, especially in the contexts of hedging or volatility prediction, nor does it correlate with alpha generation.
Furthermore, R^2 can be misleading due to its reliance on model complexity—it's inherently biased as it increases with the number of predictors. This can explain the proliferation of factors in commercial models. Additionally, data mining is a significant risk when assessing R^2, particularly given the limited historical data available for testing.
Despite these drawbacks, R^2 still provides an intuitive grasp of a factor's explanatory power. The challenge lies in rigorously linking population R^2 to genuine factor model performance across risk management and alpha generation, transcending the hurdles posed by finite sample considerations and multiple testing anomalies.
@quant_feed
👍2
Exploring the Balance Between Active and Passive Investment Strategies in Modern Markets
choffstein @ twitter
orig
There’s a growing recognition that both active and passive investing strategies influence market dynamics beyond their intended effects. We see how active strategies can set prices, while passive investments adjust based on market cap, yet liquidity issues mean these influences aren’t always in sync. This highlights the importance of recognizing how different investment vehicles interact.
Target date funds are surfacing as key players, altering stock correlations, including stock/bond relationships. It's notable that financialization in the 2000s skewed commodity correlations, raising questions about our assumptions regarding market behavior.
Transacting in financial markets inevitably creates some impact—Koijen’s research underscores this point. Additionally, non-informational flows may influence prices more significantly than previously thought. Amid this complexity, discussions around the nuances often get lost, leading to polarized views that oversimplify the intricacies of market interactions.
@quant_feed
choffstein @ twitter
orig
There’s a growing recognition that both active and passive investing strategies influence market dynamics beyond their intended effects. We see how active strategies can set prices, while passive investments adjust based on market cap, yet liquidity issues mean these influences aren’t always in sync. This highlights the importance of recognizing how different investment vehicles interact.
Target date funds are surfacing as key players, altering stock correlations, including stock/bond relationships. It's notable that financialization in the 2000s skewed commodity correlations, raising questions about our assumptions regarding market behavior.
Transacting in financial markets inevitably creates some impact—Koijen’s research underscores this point. Additionally, non-informational flows may influence prices more significantly than previously thought. Amid this complexity, discussions around the nuances often get lost, leading to polarized views that oversimplify the intricacies of market interactions.
@quant_feed
The Underestimated Challenge of Long/Short Equity Success in China: A Closer Look at Foreign PMs
systematicls @ twitter
orig
The challenge of achieving success in China’s long/short equities market is significantly underestimated. I've noticed that very few non-Chinese PMs manage to sustain long-term profitability in this space. This raises a question about potential hidden advantages held by domestic PMs, suggesting they might have exclusive access to critical data or networks, particularly regarding trading suspensions.
I’m curious if there are any academic papers I’ve overlooked beyond the standard studies, particularly focusing on the impacts of trading suspensions in China. Also interested in any relevant anecdotes from others in the field.
It's clear to me that the transparency of data sources is uneven; I suspect domestic investors tap into more nuanced information that goes beyond standard terminals like Wind. This could be a key factor in their success in L/S strategies.
@quant_feed
systematicls @ twitter
orig
The challenge of achieving success in China’s long/short equities market is significantly underestimated. I've noticed that very few non-Chinese PMs manage to sustain long-term profitability in this space. This raises a question about potential hidden advantages held by domestic PMs, suggesting they might have exclusive access to critical data or networks, particularly regarding trading suspensions.
I’m curious if there are any academic papers I’ve overlooked beyond the standard studies, particularly focusing on the impacts of trading suspensions in China. Also interested in any relevant anecdotes from others in the field.
It's clear to me that the transparency of data sources is uneven; I suspect domestic investors tap into more nuanced information that goes beyond standard terminals like Wind. This could be a key factor in their success in L/S strategies.
@quant_feed